You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@mesos.apache.org by "Dominic Gregoire (JIRA)" <ji...@apache.org> on 2017/03/20 17:43:41 UTC
[jira] [Comment Edited] (MESOS-7130) port_mapping isolator:
executor hangs when running on EC2
[ https://issues.apache.org/jira/browse/MESOS-7130?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15932515#comment-15932515 ]
Dominic Gregoire edited comment on MESOS-7130 at 3/20/17 5:42 PM:
------------------------------------------------------------------
I might have run into the same issue, using {{mesos 1.1.0}} with {{libnl 3.2.29}}, on an instance with an {{ena}} interface running kernel {{4.4.51-40.58.amzn1}}
The agent is running with these flags:
{noformat}
export MESOS_isolation=cgroups/cpu,cgroups/mem,network/port_mapping
export MESOS_containerizers=mesos
export MESOS_resources="ports:[31000-32000];ephemeral_ports:[32768-57344]"
export MESOS_ephemeral_ports_per_container=1024
{noformat}
Running spark 2.1.0 with 2 mesos containers on the same host, they can connect to each other’s block manager but can’t send traffic, it stays in their send-q.
Spark is logging:
{noformat}
17/03/19 16:54:56 INFO TransportClientFactory: Successfully created connection to ip-10-32-20-34.ec2.internal/10.32.20.34:34294 after 12 ms (0 ms spent in bootstraps)
17/03/19 16:56:56 ERROR TransportChannelHandler: Connection to ip-10-32-20-34.ec2.internal/10.32.20.34:34294 has been quiet for 120000 ms while there are outstanding requests. Assuming connection is dead;
please adjust spark.network.timeout if this is wrong.
{noformat}
I can see connections established between containers but everything stays in the send Qs:
{noformat}
[root@ip-10-32-20-34 sysctl.d]# ip netns
4602 (id: 1)
4600 (id: 0)
[root@ip-10-32-20-34 sysctl.d]# ip netns exec 4600 netstat -an
Connexions Internet actives (serveurs et établies)
Proto Recv-Q Send-Q Local Address Foreign Address State
tcp 0 0 10.32.20.34:32861 0.0.0.0:* LISTEN
tcp 0 0 0.0.0.0:33003 0.0.0.0:* LISTEN
tcp 0 0 10.32.20.34:33003 10.32.20.34:57363 ESTABLISHED
tcp 0 0 10.32.20.34:33566 10.32.20.34:34294 ESTABLISHED
tcp 0 0 10.32.20.34:33658 10.32.18.185:40600 ESTABLISHED
tcp 0 0 10.32.20.34:32832 10.32.18.185:40196 ESTABLISHED
tcp 0 0 10.32.20.34:33406 10.32.20.34:5051 ESTABLISHED
Sockets du domaine UNIX actives(serveurs et établies)
Proto RefCpt Indicatrs Type Etat I-Node Chemin
unix 2 [ ] STREAM CONNECTE 21869
unix 2 [ ] STREAM CONNECTE 20339
[root@ip-10-32-20-34 sysctl.d]# ip netns exec 4602 netstat -an
Connexions Internet actives (serveurs et établies)
Proto Recv-Q Send-Q Local Address Foreign Address State
tcp 0 0 0.0.0.0:33836 0.0.0.0:* LISTEN
tcp 0 0 10.32.20.34:34294 0.0.0.0:* LISTEN
tcp 0 24229 10.32.20.34:34294 10.32.20.34:33566 ESTABLISHED
tcp 0 0 10.32.20.34:33860 10.32.18.185:40196 ESTABLISHED
tcp 0 0 10.32.20.34:34680 10.32.18.185:40600 ESTABLISHED
tcp 0 0 10.32.20.34:34434 10.32.20.34:5051 ESTABLISHED
tcp 0 0 10.32.20.34:33836 10.32.20.34:58149 ESTABLISHED
Sockets du domaine UNIX actives(serveurs et établies)
Proto RefCpt Indicatrs Type Etat I-Node Chemin
unix 2 [ ] STREAM CONNECTE 20359
unix 2 [ ] STREAM CONNECTE 20373
{noformat}
was (Author: dgregoire):
I might have run into the same issue, using {{mesos 1.1.0}} with {{libnl 3.2.29}}, on an instance with an {{ena}} interface.
The agent is running with these flags:
{noformat}
export MESOS_isolation=cgroups/cpu,cgroups/mem,network/port_mapping
export MESOS_containerizers=mesos
export MESOS_resources="ports:[31000-32000];ephemeral_ports:[32768-57344]"
export MESOS_ephemeral_ports_per_container=1024
{noformat}
Running spark 2.1.0 with 2 mesos containers on the same host, they can connect to each other’s block manager but can’t send traffic, it stays in their send-q.
Spark is logging:
{noformat}
17/03/19 16:54:56 INFO TransportClientFactory: Successfully created connection to ip-10-32-20-34.ec2.internal/10.32.20.34:34294 after 12 ms (0 ms spent in bootstraps)
17/03/19 16:56:56 ERROR TransportChannelHandler: Connection to ip-10-32-20-34.ec2.internal/10.32.20.34:34294 has been quiet for 120000 ms while there are outstanding requests. Assuming connection is dead;
please adjust spark.network.timeout if this is wrong.
{noformat}
I can see connections established between containers but everything stays in the send Qs:
{noformat}
[root@ip-10-32-20-34 sysctl.d]# ip netns
4602 (id: 1)
4600 (id: 0)
[root@ip-10-32-20-34 sysctl.d]# ip netns exec 4600 netstat -an
Connexions Internet actives (serveurs et établies)
Proto Recv-Q Send-Q Local Address Foreign Address State
tcp 0 0 10.32.20.34:32861 0.0.0.0:* LISTEN
tcp 0 0 0.0.0.0:33003 0.0.0.0:* LISTEN
tcp 0 0 10.32.20.34:33003 10.32.20.34:57363 ESTABLISHED
tcp 0 0 10.32.20.34:33566 10.32.20.34:34294 ESTABLISHED
tcp 0 0 10.32.20.34:33658 10.32.18.185:40600 ESTABLISHED
tcp 0 0 10.32.20.34:32832 10.32.18.185:40196 ESTABLISHED
tcp 0 0 10.32.20.34:33406 10.32.20.34:5051 ESTABLISHED
Sockets du domaine UNIX actives(serveurs et établies)
Proto RefCpt Indicatrs Type Etat I-Node Chemin
unix 2 [ ] STREAM CONNECTE 21869
unix 2 [ ] STREAM CONNECTE 20339
[root@ip-10-32-20-34 sysctl.d]# ip netns exec 4602 netstat -an
Connexions Internet actives (serveurs et établies)
Proto Recv-Q Send-Q Local Address Foreign Address State
tcp 0 0 0.0.0.0:33836 0.0.0.0:* LISTEN
tcp 0 0 10.32.20.34:34294 0.0.0.0:* LISTEN
tcp 0 24229 10.32.20.34:34294 10.32.20.34:33566 ESTABLISHED
tcp 0 0 10.32.20.34:33860 10.32.18.185:40196 ESTABLISHED
tcp 0 0 10.32.20.34:34680 10.32.18.185:40600 ESTABLISHED
tcp 0 0 10.32.20.34:34434 10.32.20.34:5051 ESTABLISHED
tcp 0 0 10.32.20.34:33836 10.32.20.34:58149 ESTABLISHED
Sockets du domaine UNIX actives(serveurs et établies)
Proto RefCpt Indicatrs Type Etat I-Node Chemin
unix 2 [ ] STREAM CONNECTE 20359
unix 2 [ ] STREAM CONNECTE 20373
{noformat}
> port_mapping isolator: executor hangs when running on EC2
> ---------------------------------------------------------
>
> Key: MESOS-7130
> URL: https://issues.apache.org/jira/browse/MESOS-7130
> Project: Mesos
> Issue Type: Bug
> Components: ec2, executor
> Reporter: Pierre Cheynier
>
> Hi,
> I'm experiencing a weird issue: I'm using a CI to do testing on infrastructure automation.
> I recently activated the {{network/port_mapping}} isolator.
> I'm able to make the changes work and pass the test for bare-metal servers and virtualbox VMs using this configuration.
> But when I try on EC2 (on which my CI pipeline rely) it systematically fails to run any container.
> It appears that the sandbox is created and the port_mapping isolator seems to be OK according to the logs in stdout and stderr and the {tc} output :
> {noformat}
> + mount --make-rslave /run/netns
> + test -f /proc/sys/net/ipv6/conf/all/disable_ipv6
> + echo 1
> + ip link set lo address 02:44:20:bb:42:cf mtu 9001 up
> + ethtool -K eth0 rx off
> (...)
> + tc filter show dev eth0 parent ffff:0
> + tc filter show dev lo parent ffff:0
> I0215 16:01:13.941375 1 exec.cpp:161] Version: 1.0.2
> {noformat}
> Then the executor never come back in REGISTERED state and hang indefinitely.
> {GLOG_v=3} doesn't help here.
> My skills in this area are limited, but trying to load the symbols and attach a gdb to the mesos-executor process, I'm able to print this stack:
> {noformat}
> #0 0x00007feffc1386d5 in pthread_cond_wait@@GLIBC_2.3.2 () from /usr/lib64/libpthread.so.0
> #1 0x00007feffbed69ec in std::condition_variable::wait(std::unique_lock<std::mutex>&) () from /usr/lib64/libstdc++.so.6
> #2 0x00007ff0003dd8ec in void synchronized_wait<std::condition_variable, std::mutex>(std::condition_variable*, std::mutex*) () from /usr/lib64/libmesos-1.0.2.so
> #3 0x00007ff0017d595d in Gate::arrive(long) () from /usr/lib64/libmesos-1.0.2.so
> #4 0x00007ff0017c00ed in process::ProcessManager::wait(process::UPID const&) () from /usr/lib64/libmesos-1.0.2.so
> #5 0x00007ff0017c5c05 in process::wait(process::UPID const&, Duration const&) () from /usr/lib64/libmesos-1.0.2.so
> #6 0x00000000004ab26f in process::wait(process::ProcessBase const*, Duration const&) ()
> #7 0x00000000004a3903 in main ()
> {noformat}
> I concluded that the underlying shell script launched by the isolator or the task itself is just .. blocked. But I don't understand why.
> Here is a process tree to show that I've no task running but the executor is:
> {noformat}
> root 28420 0.8 3.0 1061420 124940 ? Ssl 17:56 0:25 /usr/sbin/mesos-slave --advertise_ip=127.0.0.1 --attributes=platform:centos;platform_major_version:7;type:base --cgroups_enable_cfs --cgroups_hierarchy=/sys/fs/cgroup --cgroups_net_cls_primary_handle=0xC370 --container_logger=org_apache_mesos_LogrotateContainerLogger --containerizers=mesos,docker --credential=file:///etc/mesos-chef/slave-credential --default_container_info={"type":"MESOS","volumes":[{"host_path":"tmp","container_path":"/tmp","mode":"RW"}]} --default_role=default --docker_registry=/usr/share/mesos/users --docker_store_dir=/var/opt/mesos/store/docker --egress_unique_flow_per_container --enforce_container_disk_quota --ephemeral_ports_per_container=128 --executor_environment_variables={"PATH":"/bin:/usr/bin:/usr/sbin","CRITEO_DC":"par","CRITEO_ENV":"prod"} --image_providers=docker --image_provisioner_backend=copy --isolation=cgroups/cpu,cgroups/mem,cgroups/net_cls,namespaces/pid,disk/du,filesystem/shared,filesystem/linux,docker/runtime,network/cni,network/port_mapping --logging_level=INFO --master=zk://mesos:test@localhost.localdomain:2181/mesos --modules=file:///etc/mesos-chef/slave-modules.json --port=5051 --recover=reconnect --resources=ports:[31000-32000];ephemeral_ports:[32768-57344] --strict --work_dir=/var/opt/mesos
> root 28484 0.0 2.3 433676 95016 ? Ssl 17:56 0:00 \_ mesos-logrotate-logger --help=false --log_filename=/var/opt/mesos/slaves/cdf94219-87b2-4af2-9f61-5697f0442915-S0/frameworks/366e8ed2-730e-4423-9324-086704d182b0-0000/executors/group_simplehttp.16f7c2ee-f3a8-11e6-be1c-0242b44d071f/runs/1d3e6b1c-cda8-47e5-92c4-a161429a7ac6/stdout --logrotate_options=rotate 5 --logrotate_path=logrotate --max_size=10MB
> root 28485 0.0 2.3 499212 94724 ? Ssl 17:56 0:00 \_ mesos-logrotate-logger --help=false --log_filename=/var/opt/mesos/slaves/cdf94219-87b2-4af2-9f61-5697f0442915-S0/frameworks/366e8ed2-730e-4423-9324-086704d182b0-0000/executors/group_simplehttp.16f7c2ee-f3a8-11e6-be1c-0242b44d071f/runs/1d3e6b1c-cda8-47e5-92c4-a161429a7ac6/stderr --logrotate_options=rotate 5 --logrotate_path=logrotate --max_size=10MB
> marathon 28487 0.0 2.4 635780 97388 ? Ssl 17:56 0:00 \_ mesos-executor --launcher_dir=/usr/libexec/mesos
> {noformat}
> If someone has a clue about the issue I could experience on EC2, I would be interested to talk...
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)